Neural stochastic differential equations for particle dispersion in large-eddy simulations of homogeneous isotropic turbulence

In dilute turbulent particle-laden flows, such as atmospheric dispersion of pollutants or virus particles, the dynamics tracer-like to low inertial particles are significantly altered by fluctuating motion carrier fluid phase. Neglecting effects velocity fluctuations on particle causes poor prediction transport and dispersion. To account for phase transport, stochastic differential equations coupled with large-eddy simulation proposed model seen particle. The drift diffusion terms in equation modeled using neural networks (“neural equations”). trained direct numerical simulations (DNS) decaying homogeneous isotropic turbulence at moderate Reynolds numbers. predictability models is assessed against DNS results through a priori analyses posteriori low-to-high Total kinetic energy under-predicted 40% no model, compared data. contrast, predictions match total within 5% data low- high-inertia particles. For matches variance uncorrelated 10% results, 60%–70% under-prediction model. It concluded that applicable flow configurations involving tracer